Text mining turns collections of unstructured text into evidence that can be searched, summarized, classified, or linked to entities and topics. A real system does not “understand” every document: its output depends on the task definition, training data, language, domain, preprocessing, and evaluation design.

Sentiment analysis, often called opinion mining, estimates the evaluative stance expressed in text. A system may output polarity labels, a score, or sentiment toward a particular entity or aspect. Those predictions are not direct measurements of a writer’s private emotional state, and a neutral document-level score can hide mixed positive and negative statements.